Vision-Based Real-Time Positioning and Autonomous Navigation System Using Mobile Robot in Indoor Environments

Abstract

Research toward unmanned mobile robot navigation has gained significant importance in the last decade due to its potential applications in the location-based services industry. The increase in construction of large space indoor buildings has made difficulty for humans to operate within such environments. In this study, a mobile robot's indoor navigation algorithm is developed with vision cameras. Using two monocular cameras (one looking forward and one looking downward), the developed algorithms make use of the salient features of the environments to estimate rotational and translational motions for real-time positioning of the mobile robot. At the same time, an algorithm based on artificial landmark recognition is developed. The artificial landmark is shaped arrow based signboards with different colors representing different paths. These algorithms are integrated into a designed framework for mobile robot real-time positioning and autonomous navigation. Experiments are performed to validate the designed system using the mobile robot PIONEER P3-AT. The developed algorithm was able to detect and extract artificial landmark information up to 3 m distance for the mobile robot guidance. Experiment results show an average error of 0.167 m deviation from the ideal path, signified the good ability and performance of the development autonomous navigation algorithm.

Visual Tracking using Invariant Feature Descriptor

“…Visual tracking algorithm is specifically designed to identify the state of a target throughout the video sequence in different types of video surveillance applications, such as robot navigation, traffic detection, sport analytics, gesture recognition and animal behavior analysis (Baltzakis et al, 2012;Diop et al, 2016;Linares-Sánchez et al, 2015;Mei and Ling, 2011;Santiago et al, 2011;Shvarts and Tamre, 2012). In the large number of diverse surveillance applications, visual tracking is generally divided into two categories, which are single-target tracking and multiple-target tracking.…”